Distributed pcfg password cracking

Abstract

In digital forensics, investigators frequently face cryptographic protection that prevents access to potentially significant evidence. Since users prefer passwords that are easy to remember, they often unwittingly follow a series of common password-creation patterns. A probabilistic context-free grammar is a mathematical model that can describe such patterns and provide a smart alternative for traditional brute-force and dictionary password guessing methods. Because more complex tasks require dividing the workload among multiple nodes, in the paper, we propose a technique for distributed cracking with probabilistic grammars. The idea is to distribute partially-generated sentential forms, which reduces the amount of data necessary to transfer through the network. By performing a series of practical experiments, we compare the technique with a naive solution and show that the proposed method is superior in many use-cases.